Describing Pulmonary Nodules Using 3D Clustering

After detecting a node (tumor) on medical images, it is necessary to determine its shape, localization and type. This is important for the choice of the type of clinical intervention and other aspects of the work of radiologists. Computed detection systems effectively locate nodes using 2D computed tomography (CT) imaging of the lungs. However, a more detailed description of the node (tumor) is still a big problem. In the framework of this work, three-dimensional clustering was performed on volumetric CT images, which give an idea of ​​the node and its structure. These materials were used to describe the development of the node in successive sections of the lung. Combined algorithms for clustering and determining the characteristics of nodes in 3D visualization. Some 3D features were applied to objects grouped by K-means CT lung imaging. This approach provides a visual study of the three-dimensional shape and location of the node. This study is mainly focused on clustering in 3D in order to obtain complex information missing from the radiologist's report. In addition, to evaluate the proposed system, we used a 3D density clustering algorithm for spatial data with the presence of noise and another 3D application - a graph. The proposed method detected a difficult case and automatically provided information about the types of nodes (globular, juxtapleural, and pleural-caudal). The algorithm is tested on standard data, Based on the proposed model, it is possible to cluster lung nodes in 3D CT and determine a set of characteristics such as shape, location, and type.